This model is implement by LBP + DBN, some preprocess is done first.
1.Be sure you install MATLAB2016a+ first. 2.Change the dataset's path in bp.m. 3.A correct rate 100% will be generate.(The fvc2004 is easy.You can test SDUMLA.) 4.You can also plot the curve in last part.
1.The first Data-set can be find here fvc-2004;You can also find it in other place. It's a open source database. 2. The second Data-Set can be find hereSDUMLA_HMT
主要实现方法为降维LBP + 深度信念网络(需要了解一些机器学习的知识)
1.安装NN工具箱(高版本matlab自带)
2.在bp.m中更改路径为自己的数据集路径
3.在bp.m中去掉注释可以看到学习曲线
4.在fvc2004数据库,匹配成功率为100%,可以使用更复杂的数据集测试,如山东大学数据挖掘实验室的SDUMLA,均为开源数据集